Related Experiment Video
Updated: Apr 16, 2026
Real-Time Cardiac Mapping with a Noninvasive Imageless Electrocardiographic Imaging System
Published on: April 11, 2025
Risk-Guided Atrial Fibrillation Screening With Artificial Intelligence-Enabled Electrocardiogram Models: A VITAL-AF
Natasha A Vedage1, Sam F Friedman2, Yuchiao Chang3
1Cardiovascular Disease Initiative, Broad Institute of MIT and Harvard, Cambridge, Massachusetts, USA; Cardiovascular Research Center, Heart and Vascular Institute, Mass General Brigham, Boston, Massachusetts, USA; Telemachus and Irene Demoulas Family Foundation Center for Cardiac Arrhythmias, Heart and Vascular Institute, Mass General Brigham, Boston, Massachusetts, USA.
Screening for atrial fibrillation (AF) using artificial intelligence (AI) risk models identified high-risk individuals, improving detection efficiency. This risk-based approach shows promise for targeted AF screening in primary care.
Area of Science:
- Cardiology
- Artificial Intelligence in Medicine
- Preventive Healthcare
Background:
- Current atrial fibrillation (AF) screening relies on age (≥65 years) with limited effectiveness.
- Earlier AF detection and intervention can improve patient outcomes.
- Novel risk stratification methods are needed to enhance screening yield.
Purpose of the Study:
- To evaluate the effectiveness of artificial intelligence (AI)-based risk models in identifying individuals who benefit from AF screening.
- To assess if AF screening yields are higher in individuals identified as high-risk by validated clinical and ECG-AI models.
- To compare the performance of different risk models in predicting 2-year incident AF.
Main Methods:
- The VITAL-AF cluster-randomized trial involved patients aged ≥65 years in primary care settings.
- Risk prediction utilized the CHARGE-AF clinical score, an ECG-AI model, and a combined CH-AI model.
- Two-year incident AF discrimination was assessed using AUROC and average precision; screening effect was evaluated across risk deciles.
Main Results:
- The CH-AI model demonstrated strong discrimination for 2-year AF risk (AUROC: 0.788).
- A significant AF screening effect was observed in the highest CH-AI risk decile (10.07/100 person-years vs. 7.76 in control, P < 0.05).
- The number needed to screen was 43 per year in the top risk decile.
Conclusions:
- ECG-based AI and clinical factors effectively identify high-risk individuals for AF screening.
- A risk-based screening strategy may increase efficiency but reduce population coverage.
- Further research is needed to optimize risk-based AF screening, considering additional clinical and system factors.
